Key Highlights
- AI adoption in marketing remains broad but shallow, with most AI use focused on collaboration and assistance rather than full automation.
- Pinterest's new AI products aim to enhance discovery and personalized shopping, signaling a shift toward multi-turn, context-aware interactions.
- OpenAI's ambitious revenue forecasts are significantly overestimated, urging marketers to treat long-term AI projections as aspirational rather than definitive.
- Claude's AI models exhibit varying values across languages and models, requiring marketers to audit tone and style in multilingual deployments.
- Microsoft's new in-house AI models promise faster, cheaper content creation tools, but brands should verify output quality and consistency post-implementation.
Welcome to Unprompted: The AI Marketing Brief, where I cut through the noise in AI news and research to show marketers what’s happening — and why it matters for your work, your team and your career.
If you're like me, you remember the bygone era of content creation when "more" was the entire strategy. Full stop. More keywords, more pages, more content shoveled out the door as fast as you could publish it. And it worked. Until it didn't. Google caught up, readers got tired of it, and "more" stopped being a growth lever and started being a liability.
We're doing it again, just with AI. OpenAI is projecting $100 billion in ad revenue by 2030 — one analyst thinks they'll miss that number by 90%. Meanwhile, Pinterest dropped four new AI products in a single announcement, and Microsoft shipped two new model variants in one blog post. Nobody's pausing to ask if marketers actually need all of it. They're just building faster, betting the curve keeps going up forever.
Except the industry's own data keeps quietly disagreeing with the industry's own pitch. Google's usage study found AI is only touching about 21% of tasks in the average job — not exactly the wholesale replacement everyone's selling. And Anthropic just admitted it doesn't fully understand why its own models behave differently depending on the language you're using.
If the SEO gold rush taught me anything, it's that the correction always comes from the numbers, not the marketing copy. So before you build your upcoming plans around this week's claims, check which ones are actually measured — and which ones are just more.
Pinterest's new AI tools for the discovery era: from ads to personalized shopping
Website: Pinterest Newsroom
Just the Facts: Pinterest announced four AI products: Business Assistant, a closed-beta AI collaborator inside Ads Manager that surfaces trend data and top-performing Pins visually; Pinterest MCP, an infrastructure layer that connects Pinterest's campaign and analytics data to third-party AI copilots and agentic tools via the Model Context Protocol; an upgraded Performance+ creative model that selects the best ad variant at the impression level, which increased click volume by 7.5% in testing; and Ask Pinterest, a separate experimental app designed for conversational, multi-step shopping decisions that draws on Pinterest's proprietary Taste Graph. Alpha MCP partners include PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean and Omnicom's Jump450. Pinterest says learnings from Ask Pinterest will inform future AI experiences in the main app.
Why It Matters to Marketers:
- Pinterest embedding its campaign data into agentic workflows — so your AI copilot can pull Pinterest performance without leaving your planning tool — is a structural change in how media management works.
- The Performance+ model doesn't just pick winning ads — it picks winning variants per impression. That's a different creative briefing problem. You need more asset variation going in, not just a stronger hero concept.
- Ask Pinterest is Pinterest betting that discovery shifts to multi-turn, context-aware interactions — not keyword queries. If that behavior scales, B2B marketers ignoring visual and conversational discovery channels are building for a search model that's already changing.
- The trend visualization and proactive mobile alerts are genuinely different from standard ads manager dashboards. Even in beta, it's worth getting access — not because it's polished, but because it tells you where Pinterest's optimization logic is headed.
If the SEO gold rush taught me anything, it's that the correction always comes from the numbers, not the marketing copy.
OpenAI's Ad Business Is on Pace to Miss Its Own Forecast By 90%, Analyst Says
Author: Trishla Ostwal
Website: Adweek
Just the Facts: OpenAI's advertising business is on pace to fall roughly 90% short of the company's own five-year revenue forecast, according to EMARKETER's analysis of the AI lab's ad business projections. OpenAI has projected $2.5 billion in ad revenue this year and $100 billion by 2030, while EMARKETER's data finds that standalone chatbots like ChatGPT, Microsoft Copilot app, Google AI Mode and Amazon Alexa for Shopping will collectively generate less than $1 billion in ad revenue this year and just $5.41 billion by 2030. OpenAI began its ad trial in February and touted its ambitious projections just two months later, a forecast that EMARKETER says assumes OpenAI will capture search ad budgets en masse, dominate a fully mature chatbot ad market and outperform every ad format in history all at once.
Why It Matters to Marketers:
- This gap mirrors a broader pattern of generative AI initiatives overpromising and underdelivering on ROI, with Gartner reporting GenAI has entered its "Trough of Disillusionment" as organizations confront governance and value-proof challenges.
- Marketers building 2026–2030 media plans around chatbot advertising as a major channel should treat vendor-supplied five-year forecasts as aspirational, not budgetary, given the scale of the gap between OpenAI's claim and outside estimates.
- Longer-term, marketing ops teams can pilot small, measurable chatbot ad tests now to build first-party performance benchmarks, rather than waiting on or over-indexing to platform-provided market-size projections.
Claude's Values Across Models and Languages
Website: Anthropic
Just the Facts: Anthropic analyzed roughly 300,000 real Claude.ai conversations to measure the values Claude expresses, compressing thousands of previously identified values into four interpretable axes: Deference vs. Caution, Warmth vs. Rigor, Depth vs. Brevity, and Candor vs. Execution. The research found that different Claude models express different value profiles — Sonnet 4.6 leans toward deference, warmth and brevity, while Opus 4.7 leans more toward caution, rigor and depth — and that these expressed values also vary across the 20 most common languages used on Claude.ai, with Claude expressing more warmth in Arabic and Hindi and more rigor in English and Russian. Anthropic states it does not yet fully understand what drives these differences and outlines plans to investigate their training-data causes, their effects on users and whether this value-profiling method can be built into future model evaluation and monitoring.
Why It Matters to Marketers:
- Teams deploying AI chatbots or content tools across regions should expect tone, warmth and directness to shift meaningfully by language and model choice, not just by translation quality — a new variable for localization QA.
- Marketers using LLMs for customer-facing multilingual communications should audit output tone per language and model version rather than assume a uniform brand voice, since the article shows values expressed can vary substantially and unpredictably; this is a longer-term governance concern as multilingual AI deployment scales.
- Marketing ops and brand teams building AI usage guidelines can adopt a similar "value-sampling" practice — periodically reviewing AI-generated customer messages across languages and model versions to catch unintended tone drift before it reaches audiences at scale.
Understanding the AI Economy
Author: Zanna Iscenko and Scott Strand
Website: Google (The Keyword)
Just the Facts: Google launched the first iteration of its AI & Economy ATLAS (Activity, Task, Landscape and Adoption Study), a large-scale, de-identified study built from 15 million aggregated human-AI interactions across the Gemini App, AI Mode and the Gemini API, spanning more than 150 countries, 140 languages, 800 occupations and 4,000 tasks. The report finds that workplace AI adoption is broad but shallow, reaching 68% of occupations representing 90% of U.S. employment while being used for only about 21% of tasks within a typical job, and that most AI use at work centers on collaboration and assistance — such as ideation, strategy and information retrieval — with less than 10% of interactions fully automating tasks. The study also finds that over 86% of AI interactions occur outside of work, that AI is being used by workers in manual and technical trades as a real-time diagnostic collaborator, and that global AI adoption tracks closely with a country's GDP per capita, with some middle-income countries in South America and the Middle East adopting at rates comparable to wealthier nations.
Why It Matters to Marketers:
- The finding that AI is used for only ~21% of tasks within a typical job suggests marketers should design AI-assisted workflows (drafting, research, ideation) around specific task-level insertion points, not wholesale role replacement, per Google's own usage data.
- The "broad but shallow" adoption pattern echoes Gartner's 2025 finding that GenAI has entered a "Trough of Disillusionment" as organizations shift from broad experimentation to narrower, provable use cases.
- Content and demand-gen teams can use the collaboration-over-automation finding to prioritize AI investment in ideation, research, and information-retrieval workflows immediately, while treating full task automation as a longer-term, lower-adoption use case worth testing cautiously.
Introducing MAI-Image-2.5-Pro and MAI-Voice-2-Flash
Website: Microsoft AI
Just the Facts: Microsoft AI announced two new in-house model variants now in public preview: MAI-Image-2.5-Pro, its highest-fidelity image model priced at $5 per 1M text input tokens, $8 per 1M image input tokens, and $106 per 1M image output tokens, and MAI-Voice-2-Flash, a faster and cheaper voice model priced at $15 per 1M characters that is twice as fast and 32% less expensive than its predecessor MAI-Voice-2. The company detailed how its in-house MAI models are already running in production across several Microsoft products, including powering Bing Image Creator entirely by default, enabling image-to-image editing in PowerPoint at up to 84% lower GPU costs than GPT-Image-2, becoming the default image-editing model in OneDrive with a 26% increase in save rates, and powering voice interactions in Dynamics 365 Contact Center with up to 89% lower GPU costs. Microsoft also noted that MAI-Transcribe-1.5 now supports Dragon Copilot's multilingual medical transcription workflow across 58 languages with a reported 50% relative reduction in transcription and language-identification error rates.
Why It Matters to Marketers:
- Marketing teams using Microsoft 365 tools like PowerPoint and OneDrive for creative production can expect faster, cheaper AI-generated and AI-edited imagery as these models roll out, which may lower the cost floor for in-house visual content creation.
- Microsoft's shift to fully in-house, non-distilled models signals a broader vendor trend toward proprietary AI infrastructure over reliance on third-party model licensing, a dynamic Gartner has flagged as central to enterprise AI vendor evaluation and lock-in risk.
- Marketers relying on Bing Image Creator, PowerPoint image tools or Dynamics 365 voice agents should test output quality and brand-voice consistency after this model swap, since underlying generation models changing can shift creative results even when the interface stays the same.
About the Author
Alexis Gajewski
Contributor / AI Expert
Alexis Gajewski is the Associate Director of Newsroom Operations and Development at EndeavorB2B, where she leads editorial strategy and AI integration across a portfolio of 80+ B2B brands and 150 editors. With 18+ years in B2B media, she is best known for building the systems, training programs, and organizational infrastructure that help editorial teams operate at a higher level — faster, smarter, and with clearer standards.
Her expertise spans the full editorial stack — from SEO, GEO, and analytics to AI literacy, content strategy, and journalistic standards — with a particular focus on translating emerging technology into practical frameworks editorial teams can actually adopt. She designs and delivers training programs that meet teams where they are and build toward where the industry is going, with a specialty in AI integration that covers everything from foundational literacy to advanced workflows and agentic applications. A frequent guest on ASBPE webinars, Alexis is a recognized voice on the intersection of journalism and AI, and she writes for marketers, editors, and authors on how to thoughtfully and strategically implement AI practices.
Resources
Quiz
Your Competitive Edge, DeliveredElevate your strategy with weekly insights from marketing leaders who are redefining engagement and growth. From campaign best practices to creative innovation and data-driven trends, MarketingEDGE delivers the ideas and inspiration you need to outperform your competition.
